Anthropic Disables Fable 5 and Mythos 5 Following US Government Order

According to Anthropic, officials believe they found a way to bypass some of the models' safety protections. After reviewing the reported issue, Anthropic said the problem appears limited and does not provide major new capabilities. The company also noted that similar results can be achieved with other AI models already available to the public.

Anthropic explained that Fable 5 went through extensive security testing before launch, involving government agencies, outside experts, and internal teams. The company says its safety measures remain effective and that no major universal method has been found to break the model's protections.

While Anthropic is following the government's order, it disagrees with the decision. The company believes that removing an AI model because of a small and limited safety issue could make it much harder for AI companies to release new advanced models in the future.